EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC 4.0

Preprocessing Choices and N-gram Representation in Indonesian Product Review Sentiment Classification

Akmal Faiz Agiputra, Jumanto Unjung

Journal of Computer Engineering · 12.11.2026 · 5 (02) · 192-205

DOI: 10.25124/cepat.v5i02.12451 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Computer Engineering, online erschienen am 12.11.2026. Verfasst von Akmal Faiz Agiputra und 1 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The large volume of unstructured Indonesian product reviews on e-commerce platforms creates the need for effective sentiment classification methods. Previous studies have shown that the effectiveness of preprocessing techniques, Term Frequency-Inverse Document Frequency (TF-IDF) n-gram representations, and classification models may vary across datasets and experimental settings. This study analyzes the interaction between preprocessing scenarios, TF-IDF feature representations, and classification models for Indonesian product review sentiment classification. Experiments were conducted on the PRDECT-ID dataset using eight preprocessing scenarios, two TF-IDF n-gram representations, and Logistic Regression and Multilayer Perceptron (MLP) classifiers, evaluated with Stratified 5-Fold Cross Validation. The best configuration, combining case folding, text normalization, stopword removal, and unigram-bigram TF-IDF features with an MLP classifier, achieved an F1-score of 95.14%. Unigram-bigram representation consistently outperformed unigram across all experimental settings. Scenarios involving stemming did not consistently outperform simpler combinations such as case folding and normalization, which achieved the highest average F1-scores (94.12%-94.44%) across most models. These findings indicate that performance is influenced by the interaction between preprocessing, feature representation, and classification models, with unigram-bigram representation providing the most consistent improvements. Overall, this study provides a systematic analysis of these interactions in Indonesian product review sentiment classification.

Zitieren

APA

Agiputra, A. F., & Unjung, J. (2026). Preprocessing Choices and N-gram Representation in Indonesian Product Review Sentiment Classification. Journal of Computer Engineering, 5(02), 192-205. https://doi.org/10.25124/cepat.v5i02.12451

Vancouver

Agiputra AF, Unjung J. Preprocessing Choices and N-gram Representation in Indonesian Product Review Sentiment Classification. Journal of Computer Engineering. 2026;5(02):192-205. doi:10.25124/cepat.v5i02.12451

BibTeX
@article{Agiputra2026,
  author = {Agiputra, Akmal Faiz and Unjung, Jumanto},
  title = {Preprocessing Choices and N-gram Representation in Indonesian Product Review Sentiment Classification},
  journal = {Journal of Computer Engineering},
  year = {2026},
  volume = {5},
  number = {02},
  pages = {192--205},
  doi = {10.25124/cepat.v5i02.12451}
}
Kostenlos starten →